1. Field of the Invention
The present invention relates to an image processing device, a method, and an image processing program to detect a three-dimensional front object as a target to be detected that exists in a front area of the driver's vehicle based on image data monitored in a front area of the driver's vehicle.
2. Description of the Related Art
Recently, various types of conventional detecting techniques have been proposed to detect one or more obstacles that exist in front of a driver's vehicle using an on-vehicle camera. For example, Japanese patent laid open publication No. JP H10-222679 has disclosed an image processing device for vehicles. The conventional image processing device is capable of monitoring the front area at different timings and obtaining front image data having front objects. The conventional image processing device then determines whether a target obstacle is a two-dimensional object or a three-dimensional object based on the monitored image data at different timings. In a concrete example, the conventional image processing device monitors the front target object at different timings T0 and T1 using an on-vehicle camera, and makes image data. The conventional image processing device further detects and uses other vehicle data such as a vehicle speed using a vehicle behavior detecting device. The conventional image processing device calculates a predicted position of the vehicle at timing T1 based on the monitored image data at timing T0 and the running condition of the vehicle under the predictive condition in which the front object is a two-dimensional object. The conventional image processing device then obtains the predicted image data at timing T1. The conventional image processing device compares the monitored image data obtained at timing T1 with the predicted image data at timing T1. When the comparison result indicates that the actual images data at timing T1 and the predicted image data at timing T1 are the same, the conventional image processing device detects that the front object is a two-dimensional object. On the other hand, when the comparison result indicates that the actual images data at timing T1 and the predicted image data at timing T1 are not the same, the conventional image processing device detects that the front object is a three-dimensional object.
However, because the conventional image processing device disclosed by the conventional technique JP H10-222679 uses the image data monitored at different timings and the predicted image, when the time interval between the different monitoring timings T0 and T1 is short, it becomes difficult to correctly detect the front object because the target object, in particular, a distant object becomes small in change in the monitored image data. On the other hand, when the time interval between the monitoring timings T0 and T1 becomes long, the magnitude in change of a near object becomes extremely large, and this imposes a heavy load on the conventional image processing device. As a result, this makes it difficult to obtain a correct detection (or judgment) result at high speed.